A minor note on CPAN::Reporter and a broader comment on the problem of unwelcome/inaccurate smoking.
On 29 Jan 2008 01:11:20 +0100, Slaven Rezic <[EMAIL PROTECTED]> wrote: > I recently turned PERL_MM_USE_DEFAULT on, to decrease the amount on > interactive questions while doing the smokes. It turns out that this > also stops CPAN::Reporter asking the question whether to send a FAIL > reports. A bad idea, I will revert to the old settings. That's an interesting use case. PERL_MM_USE_DEFAULT on, but wanting CPAN::Reporter to still prompt on every failure. I'll have to think about a sane way to specify it. (Perhaps a 'force_prompt' config?) > In my head there are some plans of an "annocpantesters" service, which On cpan-testers-discuss there has been a good deal of discussion about what I'm now calling "CPAN Testers 2.0" -- with report transport by HTTP to a centralized service. Accordingly, author notification should come from the central service, not from individual clients. That will make it easy to add per-author preferences about notification. Likewise, it should allow testers to delete a report after submission for situations like the one Andy complained about. More generally, I'd like to make the point that reports are just *data*. They are not quality, kwalitee, or anything else like it. In my opinion, the value to me as an author of getting reports back on how my tests run on a variety of platforms far outweighs the tiny annoyance of a report or two that I can't replicate or isn't "my fault". To Andy -- I think that opting out of smoke testing is throwing the baby out with the bathwater or something like that. To all -- many of the recent complaints I've seen from modules authors about CPAN Testers relate to the fact that the CPAN Testers tools have gotten much stricter about setting expectations for specification of various dependencies (and people are providing tools to help): * If you need a library, go learn how to specify it with Devel::CheckLib. * If you don't run on some OS or care to write portable code, go specify what you do support with Devel::AssertOS. * If you don't give a fig about some version of perl like 5.6.1, then specify "use 5.006002" in your PL files These are the *same* dependencies that end users will need to know. If you release code that isn't clear about these, then you are potentially *wasting people's time* figuring out why your modules don't work on their system. One of the reasons that I wrote CPAN::Reporter in the first place was that I was tired of sending bug reports one by one for modules I tried on Windows that couldn't be bothered to use File::Spec in their test files. After digging into these over and over to figure out if the code was failing or if it was the tests failing, I just wanted to automate the author pestering process. So, yes, it's annoying to get an "out of disk space" failure, but those kinds of FAIL reports really are the exception, not the rule. Regards, David
